Simple Medication Effect Analysis

Cardiac-MRI-based phenotype

source("./functions.R")

dfs <- read_csv("../data/Derived/dataframes_drugs_self_reported.csv.gz")

## Rows: 220368 Columns: 18
## ── Column specification ──────────────────────────────────────────────────────────────────────────────
## Delimiter: ","
## chr  (2): Sex, Drug
## dbl (10): SID, age_delta, Age, packyrs, alcoholintakegpd, PriorMed, bmi, sbp...
## lgl  (6): Hypertension, Obesity, Diabetes Mellitus, CAD, Hyperchol, Heart Fa...
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

fits.df <- read_csv("../data/Derived/fits.csv")

## Rows: 248 Columns: 13
## ── Column specification ──────────────────────────────────────────────────────────────────────────────
## Delimiter: ","
## chr  (2): term, Drug
## dbl (10): estimate, std.error, statistic, p.value, conf.low, conf.high, frac...
## lgl  (1): AdditionalMarkers
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

fits.glm <- readRDS("../data/Derived/fits.glm.Rds")

0.1 Characterizing age_delta quartiles

dfs %>% 
  select(SID, Drug, PriorMed) %>% 
  pivot_wider(id_cols=SID,
              names_from=Drug,
              values_from=PriorMed) %>% 
  inner_join(dfs %>% 
               filter(Drug=="Statins") %>% # for example
               select(-Drug, -PriorMed),
               by="SID"
               ) ->
  quartiles.df

(quartiles.df$age_delta %>% 
  quantile(probs=c(0,0.25, .5, .75,1)) -> age_delta.quantiles)

##          0%         25%         50%         75%        100% 
## -32.3968598  -5.2954310   0.1598716   5.3321624  31.8912520

(sprintf("(%.1f, %.1f]",
        age_delta.quantiles %>% head(-1),
        age_delta.quantiles %>% tail(-1)
        ) -> age_delta.quantiles.titles)

## [1] "(-32.4, -5.3]" "(-5.3, 0.2]"   "(0.2, 5.3]"    "(5.3, 31.9]"

age_delta.quantiles %>% tail(-1) -> age_delta.quantiles

quartiles.df %>% 
  mutate(age_delta_cat=case_when(
    age_delta <= age_delta.quantiles[1] ~ names(age_delta.quantiles)[1],
    age_delta <= age_delta.quantiles[2] ~ names(age_delta.quantiles)[2],
    age_delta <= age_delta.quantiles[3] ~ names(age_delta.quantiles)[3],
    T ~ "100%"
  )) -> 
  quartiles.df

quartiles.df %>% 
  mutate(age_delta_cat=factor(age_delta_cat,
                              levels=names(age_delta.quantiles),
                              labels=age_delta.quantiles.titles,
                              ordered=TRUE
         )) ->
  quartiles.df

quartiles.df %>%
  drop_na()%>%
  select(-SID) %>% 
  gtsummary::tbl_summary(
    by=age_delta_cat,
    label  = list(
      age_delta   ~ "Cardiac Age gap",
      CAD ~ "Coronary Artery Disease",
      Hyperchol~"Hypercholesterolemia",
      packyrs~ "Packs year of smoking",
      alcoholintakegpd~ "Alcohol is g/day consumed"
    ),
    missing_text = "Missing",
    digits = list(age_delta ~ c(1, 1),
                  packyrs ~ c(2,1),
                  alcoholintakegpd ~ c(1,1)
                  )
    
  ) %>%
  gtsummary::add_p()

Characteristic	(-32.4, -5.3], N = 6,8871	(-5.3, 0.2], N = 6,8861	(0.2, 5.3], N = 6,8861	(5.3, 31.9], N = 6,8871	p-value2
Diuretics	252 (3.7%)	316 (4.6%)	314 (4.6%)	348 (5.1%)	<0.001
CCB	543 (7.9%)	737 (11%)	742 (11%)	752 (11%)	<0.001
ARB	339 (4.9%)	383 (5.6%)	452 (6.6%)	468 (6.8%)	<0.001
ACEi	520 (7.6%)	627 (9.1%)	599 (8.7%)	666 (9.7%)	<0.001
BB	480 (7.0%)	499 (7.2%)	491 (7.1%)	448 (6.5%)	0.3
Digoxin	8 (0.1%)	8 (0.1%)	9 (0.1%)	11 (0.2%)	0.9
Metformin	189 (2.7%)	232 (3.4%)	244 (3.5%)	302 (4.4%)	<0.001
Statins	1,498 (22%)	1,686 (24%)	1,840 (27%)	1,670 (24%)	<0.001
Cardiac Age gap	-9.0 (-11.9, -7.0)	-2.5 (-3.9, -1.1)	2.6 (1.4, 3.9)	8.7 (6.9, 11.3)	<0.001
Sex					<0.001
Female	3,681 (53%)	3,477 (50%)	3,395 (49%)	3,284 (48%)
Male	3,206 (47%)	3,409 (50%)	3,491 (51%)	3,603 (52%)
Age	63 (56, 69)	65 (58, 71)	66 (60, 71)	64 (59, 69)	<0.001
Packs year of smoking	0.00 (0.0, 1.16)	0.00 (0.0, 3.00)	0.00 (0.0, 3.75)	0.00 (0.0, 4.50)	<0.001
Alcohol is g/day consumed	3.2 (0.0, 14.2)	3.8 (0.0, 15.2)	4.1 (0.0, 16.6)	3.4 (0.0, 15.3)	<0.001
Hypertension	1,742 (25%)	2,266 (33%)	2,448 (36%)	2,742 (40%)	<0.001
Obesity	1,497 (22%)	1,380 (20%)	1,416 (21%)	1,382 (20%)	0.047
Diabetes Mellitus	432 (6.3%)	474 (6.9%)	502 (7.3%)	575 (8.3%)	<0.001
Coronary Artery Disease	487 (7.1%)	549 (8.0%)	564 (8.2%)	508 (7.4%)	0.049
Hypercholesterolemia	1,418 (21%)	1,582 (23%)	1,700 (25%)	1,591 (23%)	<0.001
Heart Failure	96 (1.4%)	107 (1.6%)	97 (1.4%)	112 (1.6%)	0.6
bmi	26.1 (23.6, 29.4)	26.2 (23.7, 29.3)	26.3 (23.9, 29.2)	26.5 (24.0, 29.3)	0.003
sbp	131 (120, 143)	137 (126, 150)	140 (129, 153)	142 (131, 156)	<0.001
dbp	76 (69, 82)	78 (72, 84)	80 (73, 86)	82 (75, 89)	<0.001
pulse_rate	66 (58, 73)	67 (60, 76)	68 (61, 76)	71 (64, 79)	<0.001
1 n (%); Median (IQR)
2 Pearson's Chi-squared test; Kruskal-Wallis rank sum test

quartiles.df %>%
  select(age_delta_cat,
         Diuretics,
         CCB,
         ARB,
         ACEi,
         BB,
         Digoxin,
         Metformin,
         Statins,
         Sex,
         Age,
         age_delta
         ) %>% 
  drop_na()%>%
  gtsummary::tbl_summary(
    by=age_delta_cat,
    label  = list(
      age_delta   ~ "Cardiac Age gap"
    ),
    missing_text = "Missing",
    digits = list(age_delta ~ c(1, 1)
                  )
    
  ) %>%
  gtsummary::add_p()

Characteristic	(-32.4, -5.3], N = 6,8871	(-5.3, 0.2], N = 6,8861	(0.2, 5.3], N = 6,8861	(5.3, 31.9], N = 6,8871	p-value2
Diuretics	252 (3.7%)	316 (4.6%)	314 (4.6%)	348 (5.1%)	<0.001
CCB	543 (7.9%)	737 (11%)	742 (11%)	752 (11%)	<0.001
ARB	339 (4.9%)	383 (5.6%)	452 (6.6%)	468 (6.8%)	<0.001
ACEi	520 (7.6%)	627 (9.1%)	599 (8.7%)	666 (9.7%)	<0.001
BB	480 (7.0%)	499 (7.2%)	491 (7.1%)	448 (6.5%)	0.3
Digoxin	8 (0.1%)	8 (0.1%)	9 (0.1%)	11 (0.2%)	0.9
Metformin	189 (2.7%)	232 (3.4%)	244 (3.5%)	302 (4.4%)	<0.001
Statins	1,498 (22%)	1,686 (24%)	1,840 (27%)	1,670 (24%)	<0.001
Sex					<0.001
Female	3,681 (53%)	3,477 (50%)	3,395 (49%)	3,284 (48%)
Male	3,206 (47%)	3,409 (50%)	3,491 (51%)	3,603 (52%)
Age	63 (56, 69)	65 (58, 71)	66 (60, 71)	64 (59, 69)	<0.001
Cardiac Age gap	-9.0 (-11.9, -7.0)	-2.5 (-3.9, -1.1)	2.6 (1.4, 3.9)	8.7 (6.9, 11.3)	<0.001
1 n (%); Median (IQR)
2 Pearson's Chi-squared test; Kruskal-Wallis rank sum test

0.2 Correlation of features

quartiles.df %>% 
  select(-SID,
         -age_delta_cat
         ) %>% 
  mutate(Male=if_else(Sex=="Male", 1, 0)) %>% 
  select(-Sex) %>% 
  mutate_all(as.double) %>% 
  as.matrix()->
  X

Hmisc::rcorr(X, type="pearson") -> cor.pearson
cor.pearson$r -> M

corrplot::corrplot( M,
                   method="ellipse",
                   type="lower",
                   na.label='-', 
                   tl.cex = 0.7, 
                   order="AOE",
                   tl.col="black",
                   tl.srt=45,
                   number.cex = 0.3) 

corrplot::corrplot( M,
                   method="ellipse",
                   type="lower",
                   na.label='-', 
                   tl.cex = 0.7, 
                   order="hclust",
                   tl.col="black",
                   tl.srt=45,
                   number.cex = 0.3) 

0.3 Medication-Effect Overview

0.3.1 Forest plot

(fits.df  %>% 
  filter(term=="PriorMed") %>% 
  mutate(Adjustment=if_else(AdditionalMarkers==TRUE, "With SBP/DBP/HR", "Without SBP/DBP/HR")) %>% 
  mutate(Adjustment=factor(Adjustment, levels=c("Without SBP/DBP/HR","With SBP/DBP/HR"), ordered=TRUE)) %>% 
  ggplot(aes(x=Drug, y=estimate, ymin=conf.low, ymax=conf.high, color=Adjustment))+
  geom_point( position=position_dodge(width=0.3))+
  geom_linerange( position=position_dodge(width=0.3))+
  coord_flip()+
  xlab("Drug")+
  ylab("Coefficient")+
  geom_hline(yintercept=0, alpha=0.5, linetype="dashed")+
  scale_color_manual(values = c("Without SBP/DBP/HR" = "red",
                                    "With SBP/DBP/HR" = "forestgreen"
                                    )) -> p)

p+geom_segment(aes(y = 0.05, x = .55, yend = 3, xend = .55),
                  arrow = arrow(length = unit(0.25, "cm")), show.legend=FALSE, color="black")+
    geom_text(aes(y = 1.5, x = .725, label = "Drug increases cardiac age gap"), size=3, color="black") +
geom_segment(aes(y = -0.05, x = .55, yend = -3, xend = .55),
                  arrow = arrow(length = unit(0.25, "cm")), show.legend=FALSE, color="black")+
    geom_text(aes(y = -1.5, x = .725, label = "Drug decreases cardiac age gap"), size=3, color="black")

(fits.df  %>% 
  filter(term=="PriorMed") %>% 
  filter(AdditionalMarkers==TRUE) %>% 
  ggplot(aes(x=Drug, y=estimate, ymin=conf.low, ymax=conf.high))+
  geom_point( )+
  geom_linerange()+
  coord_flip()+
  xlab("Drug")+
  ylab("Coefficient")+
  geom_hline(yintercept=0, alpha=0.5, linetype="dashed") -> p)

fits.df  %>% 
  filter(term=="PriorMed") %>% 
  filter(AdditionalMarkers==TRUE) %>% 
  mutate(n=round(fracPriorMed*nobs)) ->
  phewas_results

plot_data <- phewas_results %>%
    dplyr::select(Drug, p.value, estimate, conf.high, conf.low, n) %>%
    dplyr::mutate(p.value.str=sprintf("%.3f", p.value), beta.str=sprintf("%.2f", estimate))

  parm <- plot_data %>% dplyr::mutate(mean=estimate,lower=conf.low,upper=conf.high) %>%
    dplyr::select(mean, lower, upper)
  parm <- rbind(rep(NA,3), parm)

  labeltext <- plot_data %>% dplyr::select(Drug, n, beta.str, p.value.str)
  if (nrow(labeltext)>1) {
    labeltext <- as.matrix(sapply(labeltext, as.character))
  }
  labeltext <- rbind(c('Drug','N', 'Coefficient', 'P-Value'), labeltext)

fig_width <- 8
forestplot::forestplot(labeltext=labeltext, 
                       parm, 
                       zero=0, 
                       xlab='Coefficient',
                       title = "Medication Effects", 
                       is.summary=c(TRUE, rep(FALSE, nrow(plot_data))) ,
                       txt_gp=forestplot::fpTxtGp(ticks=gpar(cex=1.2),xlab=gpar(cex=1.5), title = gpar(cex=2.0), cex=1.2), 
                       graph.pos = 4,
                       graphwidth = grid::unit(fig_width,"cm"), 
                       lwd.zero=3, 
                       lwd.ci=3)

0.3.2 Table

fits.df  %>% 
  filter(term=="PriorMed") %>% 
  mutate(Adjustment=if_else(AdditionalMarkers==TRUE, "With SBP/DBP/HR", "Without SBP/DBP/HR")) %>% 
  mutate(Adjustment=factor(Adjustment, levels=c("Without SBP/DBP/HR","With SBP/DBP/HR"), ordered=TRUE)) %>% 
  transmute(Drug,Adjustment, estimate, std.error, statistic, p.value, conf.low, conf.high) %>% 
  mutate(across(.cols=estimate:conf.high, ~round(.,2)))->table1

table1 %>% 
  kbl(caption = "Table S1: Coefficients of self-reported drugs") %>%
  kable_classic(full_width = F, html_font = "Cambria")

Table S1: Coefficients of self-reported drugs

Drug	Adjustment	estimate	std.error	statistic	p.value	conf.low	conf.high
Diuretics	With SBP/DBP/HR	0.11	0.22	0.51	0.61	-0.32	0.55
Diuretics	Without SBP/DBP/HR	-0.29	0.23	-1.25	0.21	-0.74	0.16
CCB	With SBP/DBP/HR	-0.34	0.16	-2.08	0.04	-0.66	-0.02
CCB	Without SBP/DBP/HR	-0.69	0.17	-4.05	0.00	-1.02	-0.35
ARB	With SBP/DBP/HR	0.18	0.20	0.94	0.35	-0.20	0.57
ARB	Without SBP/DBP/HR	-0.06	0.20	-0.29	0.77	-0.46	0.34
ACEi	With SBP/DBP/HR	-0.04	0.17	-0.26	0.79	-0.37	0.29
ACEi	Without SBP/DBP/HR	-0.42	0.18	-2.39	0.02	-0.76	-0.08
BB	With SBP/DBP/HR	0.61	0.19	3.18	0.00	0.23	0.98
BB	Without SBP/DBP/HR	-0.82	0.20	-4.17	0.00	-1.20	-0.43
Digoxin	With SBP/DBP/HR	0.55	1.21	0.45	0.65	-1.83	2.92
Digoxin	Without SBP/DBP/HR	0.82	1.27	0.64	0.52	-1.67	3.30
Metformin	With SBP/DBP/HR	0.80	0.33	2.46	0.01	0.16	1.44
Metformin	Without SBP/DBP/HR	0.95	0.34	2.80	0.01	0.29	1.62
Statins	With SBP/DBP/HR	0.11	0.14	0.82	0.41	-0.16	0.38
Statins	Without SBP/DBP/HR	-0.25	0.14	-1.78	0.07	-0.53	0.03

fits.df  %>% 
  filter(term=="PriorMed") %>% 
  filter(AdditionalMarkers==TRUE) %>% 
  select(-AdditionalMarkers) %>% 
  transmute(Drug, estimate, std.error, statistic, p.value, conf.low, conf.high) %>% 
  mutate(across(.cols=estimate:conf.high, ~round(.,2))) -> table2

table2 %>% 
  kbl(caption = "Table 1: Coefficients of self-reported drugs") %>%
  kable_classic(full_width = F, html_font = "Cambria")

Table 1: Coefficients of self-reported drugs

Drug	estimate	std.error	statistic	p.value	conf.low	conf.high
Diuretics	0.11	0.22	0.51	0.61	-0.32	0.55
CCB	-0.34	0.16	-2.08	0.04	-0.66	-0.02
ARB	0.18	0.20	0.94	0.35	-0.20	0.57
ACEi	-0.04	0.17	-0.26	0.79	-0.37	0.29
BB	0.61	0.19	3.18	0.00	0.23	0.98
Digoxin	0.55	1.21	0.45	0.65	-1.83	2.92
Metformin	0.80	0.33	2.46	0.01	0.16	1.44
Statins	0.11	0.14	0.82	0.41	-0.16	0.38

0.4 Statins

statin.df <- dfs %>% 
  filter(Drug=="Statins") %>% 
  select(-Drug)

statin.fits.df <- fits.df %>% 
  filter(Drug=="Statins")

get_pretty_fit_table(statin.fits.df, "Statins")

Statins: Coefficients from linear models

term	estimate	std.error	statistic	p.value	conf.low	conf.high	Adjustment
(Intercept)	-20.526	0.517	-39.703	0.000	-21.539	-19.512	With SBP/DBP/HR
(Intercept)	-0.876	0.359	-2.437	0.015	-1.580	-0.171	Without SBP/DBP/HR
Diabetes MellitusTRUE	0.645	0.178	3.628	0.000	0.297	0.994	With SBP/DBP/HR
Diabetes MellitusTRUE	0.649	0.185	3.510	0.000	0.287	1.012	Without SBP/DBP/HR
Heart FailureTRUE	0.107	0.370	0.288	0.773	-0.619	0.832	With SBP/DBP/HR
Heart FailureTRUE	-0.125	0.387	-0.323	0.746	-0.885	0.634	Without SBP/DBP/HR
alcoholintakegpd	0.001	0.003	0.328	0.743	-0.004	0.006	With SBP/DBP/HR
alcoholintakegpd	0.006	0.003	2.325	0.020	0.001	0.011	Without SBP/DBP/HR
bmi	-0.147	0.014	-10.553	0.000	-0.175	-0.120	With SBP/DBP/HR
bmi	0.000	0.014	-0.002	0.999	-0.028	0.028	Without SBP/DBP/HR
CADTRUE	0.438	0.178	2.461	0.014	0.089	0.786	With SBP/DBP/HR
CADTRUE	-0.440	0.185	-2.381	0.017	-0.803	-0.078	Without SBP/DBP/HR
dbp	0.087	0.006	14.002	0.000	0.075	0.100	With SBP/DBP/HR
HypercholTRUE	-0.198	0.132	-1.497	0.134	-0.458	0.061	With SBP/DBP/HR
HypercholTRUE	-0.095	0.138	-0.684	0.494	-0.366	0.177	Without SBP/DBP/HR
HypertensionTRUE	0.646	0.103	6.298	0.000	0.445	0.846	With SBP/DBP/HR
HypertensionTRUE	1.838	0.104	17.683	0.000	1.634	2.042	Without SBP/DBP/HR
ObesityTRUE	-0.255	0.148	-1.722	0.085	-0.546	0.035	With SBP/DBP/HR
ObesityTRUE	-0.741	0.155	-4.786	0.000	-1.044	-0.437	Without SBP/DBP/HR
packyrs	0.023	0.004	6.029	0.000	0.015	0.030	With SBP/DBP/HR
packyrs	0.022	0.004	5.499	0.000	0.014	0.030	Without SBP/DBP/HR
poly(Age, 2)1	-69.744	8.444	-8.260	0.000	-86.294	-53.194	With SBP/DBP/HR
poly(Age, 2)1	-22.395	8.031	-2.788	0.005	-38.137	-6.653	Without SBP/DBP/HR
poly(Age, 2)2	-106.586	7.248	-14.705	0.000	-120.793	-92.378	With SBP/DBP/HR
poly(Age, 2)2	-112.960	7.565	-14.932	0.000	-127.787	-98.132	Without SBP/DBP/HR
PriorMed	0.112	0.137	0.818	0.414	-0.156	0.380	With SBP/DBP/HR
PriorMed	-0.254	0.143	-1.781	0.075	-0.534	0.026	Without SBP/DBP/HR
pulse_rate	0.110	0.004	27.855	0.000	0.103	0.118	With SBP/DBP/HR
sbp	0.067	0.003	19.643	0.000	0.061	0.074	With SBP/DBP/HR
SexMale	0.442	0.095	4.648	0.000	0.256	0.629	With SBP/DBP/HR
SexMale	0.540	0.096	5.593	0.000	0.351	0.729	Without SBP/DBP/HR

get_pretty_fit_forest_plot(statin.fits.df) ->p
p

statin.fits.df %>% 
  get_fit_glance

generate_table.self.reported(statin.df, "Statins")

Characteristic	No Statins, N = 20,8521	Statins, N = 6,6941	p-value2
Cardiac Age gap	0.0 (-5.5, 5.3)	0.6 (-4.6, 5.3)	<0.001
Sex			<0.001
Female	11,672 (56%)	2,165 (32%)
Male	9,180 (44%)	4,529 (68%)
Age	63 (57, 68)	69 (64, 72)	<0.001
Packs year of smoking	0 (0, 1)	0 (0, 8)	<0.001
Alcohol is g/day consumed	3 (0, 14)	4 (0, 18)	<0.001
Hypertension	5,419 (26%)	3,779 (56%)	<0.001
Obesity	3,889 (19%)	1,786 (27%)	<0.001
Diabetes Mellitus	770 (3.7%)	1,213 (18%)	<0.001
Coronary Artery Disease	727 (3.5%)	1,381 (21%)	<0.001
Hypercholesterolemia	1,711 (8.2%)	4,580 (68%)	<0.001
Heart Failure	190 (0.9%)	222 (3.3%)	<0.001
dbp	78 (72, 86)	78 (72, 85)	0.12
sbp	136 (125, 150)	142 (130, 154)	<0.001
pulse_rate	68 (61, 76)	68 (60, 76)	0.054
bmi	25.9 (23.5, 28.9)	27.3 (24.9, 30.3)	<0.001
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Pearson's Chi-squared test

statin.df %>% 
  get_stratified_age_gap_histogram("Statins")

statin.df %>% 
  get_stratified_age_gap_quantiles("Statins")

0.4.1 Lasso

statin.glmnet.cv <- fits.glm[["Statins"]]

plot(statin.glmnet.cv) 

coef(statin.glmnet.cv, c(statin.glmnet.cv$lambda.min,
                   statin.glmnet.cv$lambda.1se))

## 17 x 2 sparse Matrix of class "dgCMatrix"
##                                    s1            s2
## (Intercept)             -2.275099e-16 -1.080653e-16
## SID                      6.956339e-04  .           
## SexMale                  2.260680e-02  .           
## Age                     -4.736318e-02  .           
## packyrs                  3.460619e-02  7.207680e-03
## alcoholintakegpd         2.768395e-03  .           
## PriorMed                 .             .           
## HypertensionTRUE         3.887463e-02  1.273594e-02
## ObesityTRUE             -9.817050e-03  .           
## `Diabetes Mellitus`TRUE  1.989265e-02  .           
## CADTRUE                  1.123753e-02  .           
## HypercholTRUE           -5.070867e-03  .           
## `Heart Failure`TRUE      .             .           
## bmi                     -8.337809e-02 -3.636707e-02
## sbp                      1.542788e-01  1.182037e-01
## dbp                      1.232901e-01  1.244400e-01
## pulse_rate               1.640762e-01  1.306085e-01

0.5 Metformin

metformin.df <- dfs %>% 
  filter(Drug=="Metformin") %>% 
  select(-Drug)

metformin.fits.df <- fits.df %>% 
  filter(Drug=="Metformin")

get_pretty_fit_table(metformin.fits.df, "Metformin")

Metformin: Coefficients from linear models

term	estimate	std.error	statistic	p.value	conf.low	conf.high	Adjustment
(Intercept)	-20.492	0.517	-39.642	0.000	-21.505	-19.479	With SBP/DBP/HR
(Intercept)	-0.839	0.359	-2.336	0.019	-1.543	-0.135	Without SBP/DBP/HR
Diabetes MellitusTRUE	0.286	0.234	1.218	0.223	-0.174	0.745	With SBP/DBP/HR
Diabetes MellitusTRUE	0.150	0.245	0.612	0.541	-0.330	0.630	Without SBP/DBP/HR
Heart FailureTRUE	0.114	0.370	0.308	0.758	-0.612	0.840	With SBP/DBP/HR
Heart FailureTRUE	-0.117	0.387	-0.302	0.763	-0.876	0.642	Without SBP/DBP/HR
alcoholintakegpd	0.001	0.003	0.376	0.707	-0.004	0.006	With SBP/DBP/HR
alcoholintakegpd	0.006	0.003	2.389	0.017	0.001	0.012	Without SBP/DBP/HR
bmi	-0.147	0.014	-10.550	0.000	-0.174	-0.120	With SBP/DBP/HR
bmi	-0.001	0.014	-0.104	0.918	-0.029	0.026	Without SBP/DBP/HR
CADTRUE	0.461	0.176	2.615	0.009	0.116	0.807	With SBP/DBP/HR
CADTRUE	-0.478	0.183	-2.611	0.009	-0.838	-0.119	Without SBP/DBP/HR
dbp	0.088	0.006	14.052	0.000	0.075	0.100	With SBP/DBP/HR
HypercholTRUE	-0.147	0.112	-1.317	0.188	-0.367	0.072	With SBP/DBP/HR
HypercholTRUE	-0.236	0.117	-2.019	0.044	-0.465	-0.007	Without SBP/DBP/HR
HypertensionTRUE	0.652	0.102	6.405	0.000	0.453	0.852	With SBP/DBP/HR
HypertensionTRUE	1.815	0.103	17.561	0.000	1.612	2.018	Without SBP/DBP/HR
ObesityTRUE	-0.270	0.148	-1.818	0.069	-0.560	0.021	With SBP/DBP/HR
ObesityTRUE	-0.753	0.155	-4.863	0.000	-1.056	-0.449	Without SBP/DBP/HR
packyrs	0.023	0.004	6.018	0.000	0.015	0.030	With SBP/DBP/HR
packyrs	0.021	0.004	5.418	0.000	0.014	0.029	Without SBP/DBP/HR
poly(Age, 2)1	-68.470	8.356	-8.194	0.000	-84.849	-52.091	With SBP/DBP/HR
poly(Age, 2)1	-24.417	7.936	-3.077	0.002	-39.971	-8.862	Without SBP/DBP/HR
poly(Age, 2)2	-106.349	7.246	-14.676	0.000	-120.552	-92.146	With SBP/DBP/HR
poly(Age, 2)2	-113.168	7.562	-14.965	0.000	-127.991	-98.346	Without SBP/DBP/HR
PriorMed	0.799	0.325	2.456	0.014	0.161	1.437	With SBP/DBP/HR
PriorMed	0.952	0.340	2.799	0.005	0.285	1.618	Without SBP/DBP/HR
pulse_rate	0.110	0.004	27.703	0.000	0.102	0.118	With SBP/DBP/HR
sbp	0.067	0.003	19.608	0.000	0.061	0.074	With SBP/DBP/HR
SexMale	0.442	0.095	4.668	0.000	0.256	0.628	With SBP/DBP/HR
SexMale	0.515	0.096	5.370	0.000	0.327	0.704	Without SBP/DBP/HR

get_pretty_fit_forest_plot(metformin.fits.df) ->p
p

metformin.fits.df %>% 
  get_fit_glance

generate_table.self.reported(metformin.df, "Metformin")

Characteristic	Metformin, N = 9671	No Metformin, N = 26,5791	p-value2
Cardiac Age gap	1.4 (-3.7, 6.8)	0.1 (-5.4, 5.3)	<0.001
Sex			<0.001
Female	299 (31%)	13,538 (51%)
Male	668 (69%)	13,041 (49%)
Age	66 (60, 71)	64 (58, 70)	<0.001
Packs year of smoking	0 (0, 13)	0 (0, 3)	<0.001
Alcohol is g/day consumed	1 (0, 11)	4 (0, 15)	<0.001
Hypertension	630 (65%)	8,568 (32%)	<0.001
Obesity	541 (56%)	5,134 (19%)	<0.001
Diabetes Mellitus	967 (100%)	1,016 (3.8%)	<0.001
Coronary Artery Disease	174 (18%)	1,934 (7.3%)	<0.001
Hypercholesterolemia	516 (53%)	5,775 (22%)	<0.001
Heart Failure	38 (3.9%)	374 (1.4%)	<0.001
dbp	78 (71, 84)	78 (72, 86)	0.005
sbp	140 (130, 153)	138 (126, 150)	<0.001
pulse_rate	73 (65, 83)	68 (60, 76)	<0.001
bmi	29.7 (26.7, 33.5)	26.2 (23.7, 29.1)	<0.001
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Pearson's Chi-squared test

metformin.df %>% 
  get_stratified_age_gap_histogram("Metformin")

metformin.df %>% 
    get_stratified_age_gap_quantiles("Metformin")

0.5.1 Lasso

metformin.glmnet.cv <- fits.glm[["Metformin"]]

plot(metformin.glmnet.cv) 

coef(metformin.glmnet.cv, c(metformin.glmnet.cv$lambda.min,
                   metformin.glmnet.cv$lambda.1se))

## 17 x 2 sparse Matrix of class "dgCMatrix"
##                                    s1            s2
## (Intercept)             -1.919073e-16 -1.144220e-16
## SID                      1.997608e-03  .           
## SexMale                  2.326247e-02  .           
## Age                     -4.960966e-02 -3.901188e-04
## packyrs                  3.571725e-02  1.000863e-02
## alcoholintakegpd         3.916300e-03  .           
## PriorMed                 1.948702e-02  .           
## HypertensionTRUE         4.019460e-02  1.517539e-02
## ObesityTRUE             -1.121007e-02  .           
## `Diabetes Mellitus`TRUE  8.450938e-03  .           
## CADTRUE                  1.318581e-02  .           
## HypercholTRUE           -7.544698e-03  .           
## `Heart Failure`TRUE      1.049425e-04  .           
## bmi                     -8.587474e-02 -4.076007e-02
## sbp                      1.559864e-01  1.195392e-01
## dbp                      1.237520e-01  1.262449e-01
## pulse_rate               1.652048e-01  1.331395e-01

0.6 Digoxin

digoxin.df <- dfs %>% 
  filter(Drug=="Digoxin") %>% 
  select(-Drug)

digoxin.fits.df <- fits.df %>% 
  filter(Drug=="Digoxin")

get_pretty_fit_table(digoxin.fits.df, "Digoxin")

Digoxin: Coefficients from linear models

term	estimate	std.error	statistic	p.value	conf.low	conf.high	Adjustment
(Intercept)	-20.518	0.517	-39.695	0.000	-21.531	-19.504	With SBP/DBP/HR
(Intercept)	-0.858	0.359	-2.390	0.017	-1.562	-0.154	Without SBP/DBP/HR
Diabetes MellitusTRUE	0.662	0.176	3.755	0.000	0.317	1.008	With SBP/DBP/HR
Diabetes MellitusTRUE	0.601	0.183	3.276	0.001	0.241	0.960	Without SBP/DBP/HR
Heart FailureTRUE	0.088	0.372	0.237	0.813	-0.642	0.818	With SBP/DBP/HR
Heart FailureTRUE	-0.152	0.390	-0.391	0.696	-0.916	0.611	Without SBP/DBP/HR
alcoholintakegpd	0.001	0.003	0.323	0.747	-0.004	0.006	With SBP/DBP/HR
alcoholintakegpd	0.006	0.003	2.328	0.020	0.001	0.011	Without SBP/DBP/HR
bmi	-0.147	0.014	-10.522	0.000	-0.174	-0.119	With SBP/DBP/HR
bmi	-0.001	0.014	-0.072	0.942	-0.029	0.027	Without SBP/DBP/HR
CADTRUE	0.458	0.176	2.594	0.009	0.112	0.803	With SBP/DBP/HR
CADTRUE	-0.483	0.183	-2.637	0.008	-0.842	-0.124	Without SBP/DBP/HR
dbp	0.087	0.006	13.981	0.000	0.075	0.099	With SBP/DBP/HR
HypercholTRUE	-0.140	0.112	-1.250	0.211	-0.359	0.079	With SBP/DBP/HR
HypercholTRUE	-0.226	0.117	-1.936	0.053	-0.456	0.003	Without SBP/DBP/HR
HypertensionTRUE	0.655	0.102	6.429	0.000	0.455	0.854	With SBP/DBP/HR
HypertensionTRUE	1.818	0.103	17.584	0.000	1.615	2.020	Without SBP/DBP/HR
ObesityTRUE	-0.257	0.148	-1.737	0.082	-0.548	0.033	With SBP/DBP/HR
ObesityTRUE	-0.738	0.155	-4.771	0.000	-1.042	-0.435	Without SBP/DBP/HR
packyrs	0.023	0.004	6.047	0.000	0.015	0.030	With SBP/DBP/HR
packyrs	0.022	0.004	5.458	0.000	0.014	0.029	Without SBP/DBP/HR
poly(Age, 2)1	-68.835	8.358	-8.235	0.000	-85.218	-52.452	With SBP/DBP/HR
poly(Age, 2)1	-24.680	7.938	-3.109	0.002	-40.238	-9.122	Without SBP/DBP/HR
poly(Age, 2)2	-106.512	7.248	-14.696	0.000	-120.718	-92.306	With SBP/DBP/HR
poly(Age, 2)2	-113.329	7.564	-14.983	0.000	-128.155	-98.503	Without SBP/DBP/HR
PriorMed	0.550	1.212	0.454	0.650	-1.825	2.924	With SBP/DBP/HR
PriorMed	0.815	1.268	0.643	0.520	-1.669	3.300	Without SBP/DBP/HR
pulse_rate	0.110	0.004	27.838	0.000	0.102	0.118	With SBP/DBP/HR
sbp	0.067	0.003	19.635	0.000	0.061	0.074	With SBP/DBP/HR
SexMale	0.450	0.095	4.754	0.000	0.264	0.635	With SBP/DBP/HR
SexMale	0.521	0.096	5.433	0.000	0.333	0.710	Without SBP/DBP/HR

get_pretty_fit_forest_plot(digoxin.fits.df) ->p
p

digoxin.fits.df %>% 
  get_fit_glance

generate_table.self.reported(digoxin.df, "Digoxin")

Characteristic	Digoxin, N = 361	No Digoxin, N = 27,5101	p-value2
Cardiac Age gap	1.5 (-4.9, 6.7)	0.2 (-5.3, 5.3)	0.5
Sex			<0.001
Female	8 (22%)	13,829 (50%)
Male	28 (78%)	13,681 (50%)
Age	71 (66, 75)	65 (58, 70)	<0.001
Packs year of smoking	0 (0, 8)	0 (0, 3)	0.6
Alcohol is g/day consumed	2 (0, 17)	4 (0, 15)	0.7
Hypertension	23 (64%)	9,175 (33%)	<0.001
Obesity	12 (33%)	5,663 (21%)	0.059
Diabetes Mellitus	13 (36%)	1,970 (7.2%)	<0.001
Coronary Artery Disease	8 (22%)	2,100 (7.6%)	0.005
Hypercholesterolemia	12 (33%)	6,279 (23%)	0.13
Heart Failure	14 (39%)	398 (1.4%)	<0.001
dbp	77 (71, 88)	78 (72, 86)	>0.9
sbp	136 (124, 156)	138 (126, 150)	0.8
pulse_rate	74 (66, 82)	68 (61, 76)	0.006
bmi	27.3 (23.9, 31.7)	26.3 (23.8, 29.3)	0.3
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Pearson's Chi-squared test; Fisher's exact test

digoxin.df %>% 
  get_stratified_age_gap_histogram("Digoxin")

digoxin.df %>% 
  get_stratified_age_gap_quantiles("Digoxin")

0.6.1 Lasso

digoxin.glmnet.cv <- fits.glm[["Digoxin"]]

plot(digoxin.glmnet.cv) 

coef(digoxin.glmnet.cv, c(digoxin.glmnet.cv$lambda.min,
                   digoxin.glmnet.cv$lambda.1se))

## 17 x 2 sparse Matrix of class "dgCMatrix"
##                                    s1            s2
## (Intercept)             -2.330190e-16 -1.012851e-16
## SID                      1.638825e-03  .           
## SexMale                  2.340433e-02  .           
## Age                     -4.899890e-02  .           
## packyrs                  3.546793e-02  4.173688e-03
## alcoholintakegpd         3.335993e-03  .           
## PriorMed                 2.639928e-04  .           
## HypertensionTRUE         3.987529e-02  1.019900e-02
## ObesityTRUE             -1.031775e-02  .           
## `Diabetes Mellitus`TRUE  2.092399e-02  .           
## CADTRUE                  1.241705e-02  .           
## HypercholTRUE           -6.457367e-03  .           
## `Heart Failure`TRUE      .             .           
## bmi                     -8.489514e-02 -3.159110e-02
## sbp                      1.555273e-01  1.167419e-01
## dbp                      1.232406e-01  1.224064e-01
## pulse_rate               1.652914e-01  1.278304e-01

0.7 Beta Blocker

BB.df <- dfs %>% 
  filter(Drug=="BB") %>% 
  select(-Drug)

BB.fits.df <- fits.df %>% 
  filter(Drug=="BB")

get_pretty_fit_table(BB.fits.df, "Beta Blocker")

Beta Blocker: Coefficients from linear models

term	estimate	std.error	statistic	p.value	conf.low	conf.high	Adjustment
(Intercept)	-20.673	0.519	-39.825	0.000	-21.690	-19.655	With SBP/DBP/HR
(Intercept)	-0.893	0.359	-2.486	0.013	-1.596	-0.189	Without SBP/DBP/HR
Diabetes MellitusTRUE	0.649	0.176	3.681	0.000	0.304	0.995	With SBP/DBP/HR
Diabetes MellitusTRUE	0.620	0.183	3.385	0.001	0.261	0.980	Without SBP/DBP/HR
Heart FailureTRUE	0.015	0.371	0.041	0.967	-0.713	0.743	With SBP/DBP/HR
Heart FailureTRUE	-0.001	0.388	-0.002	0.999	-0.762	0.761	Without SBP/DBP/HR
alcoholintakegpd	0.001	0.003	0.293	0.770	-0.004	0.006	With SBP/DBP/HR
alcoholintakegpd	0.006	0.003	2.355	0.019	0.001	0.011	Without SBP/DBP/HR
bmi	-0.149	0.014	-10.691	0.000	-0.176	-0.122	With SBP/DBP/HR
bmi	0.001	0.014	0.046	0.963	-0.027	0.029	Without SBP/DBP/HR
CADTRUE	0.281	0.185	1.524	0.128	-0.081	0.643	With SBP/DBP/HR
CADTRUE	-0.235	0.193	-1.219	0.223	-0.613	0.143	Without SBP/DBP/HR
dbp	0.087	0.006	14.008	0.000	0.075	0.100	With SBP/DBP/HR
HypercholTRUE	-0.158	0.112	-1.416	0.157	-0.378	0.061	With SBP/DBP/HR
HypercholTRUE	-0.203	0.117	-1.733	0.083	-0.432	0.027	Without SBP/DBP/HR
HypertensionTRUE	0.609	0.103	5.925	0.000	0.408	0.811	With SBP/DBP/HR
HypertensionTRUE	1.868	0.104	17.961	0.000	1.664	2.072	Without SBP/DBP/HR
ObesityTRUE	-0.256	0.148	-1.728	0.084	-0.547	0.034	With SBP/DBP/HR
ObesityTRUE	-0.734	0.155	-4.744	0.000	-1.037	-0.431	Without SBP/DBP/HR
packyrs	0.023	0.004	6.047	0.000	0.015	0.030	With SBP/DBP/HR
packyrs	0.022	0.004	5.450	0.000	0.014	0.029	Without SBP/DBP/HR
poly(Age, 2)1	-70.335	8.370	-8.404	0.000	-86.740	-53.930	With SBP/DBP/HR
poly(Age, 2)1	-23.089	7.942	-2.907	0.004	-38.656	-7.522	Without SBP/DBP/HR
poly(Age, 2)2	-106.876	7.247	-14.748	0.000	-121.080	-92.672	With SBP/DBP/HR
poly(Age, 2)2	-112.698	7.562	-14.903	0.000	-127.520	-97.876	Without SBP/DBP/HR
PriorMed	0.606	0.190	3.182	0.001	0.233	0.979	With SBP/DBP/HR
PriorMed	-0.818	0.196	-4.165	0.000	-1.203	-0.433	Without SBP/DBP/HR
pulse_rate	0.112	0.004	28.017	0.000	0.104	0.120	With SBP/DBP/HR
sbp	0.068	0.003	19.772	0.000	0.061	0.075	With SBP/DBP/HR
SexMale	0.455	0.095	4.806	0.000	0.269	0.640	With SBP/DBP/HR
SexMale	0.523	0.096	5.450	0.000	0.335	0.711	Without SBP/DBP/HR

get_pretty_fit_forest_plot(BB.fits.df) ->p
p

BB.fits.df %>% 
  get_fit_glance

generate_table.self.reported(BB.df, "BB")

Characteristic	BB, N = 1,9181	No BB, N = 25,6281	p-value2
Cardiac Age gap	0.0 (-5.3, 5.0)	0.2 (-5.3, 5.4)	0.2
Sex			<0.001
Female	681 (36%)	13,156 (51%)
Male	1,237 (64%)	12,472 (49%)
Age	69 (63, 72)	64 (58, 69)	<0.001
Packs year of smoking	0 (0, 8)	0 (0, 3)	<0.001
Alcohol is g/day consumed	4 (0, 17)	4 (0, 15)	0.3
Hypertension	1,262 (66%)	7,936 (31%)	<0.001
Obesity	587 (31%)	5,088 (20%)	<0.001
Diabetes Mellitus	315 (16%)	1,668 (6.5%)	<0.001
Coronary Artery Disease	843 (44%)	1,265 (4.9%)	<0.001
Hypercholesterolemia	935 (49%)	5,356 (21%)	<0.001
Heart Failure	171 (8.9%)	241 (0.9%)	<0.001
dbp	76 (70, 84)	79 (72, 86)	<0.001
sbp	138 (126, 152)	138 (126, 150)	0.12
pulse_rate	61 (55, 69)	68 (62, 76)	<0.001
bmi	27.7 (25.1, 30.9)	26.2 (23.7, 29.1)	<0.001
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Pearson's Chi-squared test

BB.df %>% 
 get_stratified_age_gap_histogram("Beta Blockers")

BB.df %>% 
 get_stratified_age_gap_quantiles("Beta Blockers")

0.7.1 Lasso

BB.glmnet.cv <- fits.glm[["BB"]]

plot(BB.glmnet.cv) 

coef(BB.glmnet.cv, c(BB.glmnet.cv$lambda.min,
                   BB.glmnet.cv$lambda.1se))

## 17 x 2 sparse Matrix of class "dgCMatrix"
##                                    s1            s2
## (Intercept)             -2.424344e-16 -1.012851e-16
## SID                      1.982979e-03  .           
## SexMale                  2.395217e-02  .           
## Age                     -5.080671e-02  .           
## packyrs                  3.581281e-02  4.173688e-03
## alcoholintakegpd         3.411826e-03  .           
## PriorMed                 1.750507e-02  .           
## HypertensionTRUE         3.782778e-02  1.019900e-02
## ObesityTRUE             -1.044448e-02  .           
## `Diabetes Mellitus`TRUE  2.086409e-02  .           
## CADTRUE                  7.322854e-03  .           
## HypercholTRUE           -7.898300e-03  .           
## `Heart Failure`TRUE      .             .           
## bmi                     -8.685689e-02 -3.159110e-02
## sbp                      1.572566e-01  1.167419e-01
## dbp                      1.233614e-01  1.224064e-01
## pulse_rate               1.682552e-01  1.278304e-01

0.8 ACEi

ACEi.df <- dfs %>% 
  filter(Drug=="ACEi") %>% 
  select(-Drug)

ACEi.fits.df <- fits.df %>% 
  filter(Drug=="ACEi")

get_pretty_fit_table(ACEi.fits.df, "ACEi")

ACEi: Coefficients from linear models

term	estimate	std.error	statistic	p.value	conf.low	conf.high	Adjustment
(Intercept)	-20.514	0.517	-39.675	0.000	-21.527	-19.501	With SBP/DBP/HR
(Intercept)	-0.867	0.359	-2.415	0.016	-1.571	-0.164	Without SBP/DBP/HR
Diabetes MellitusTRUE	0.667	0.177	3.780	0.000	0.321	1.013	With SBP/DBP/HR
Diabetes MellitusTRUE	0.629	0.184	3.428	0.001	0.269	0.989	Without SBP/DBP/HR
Heart FailureTRUE	0.112	0.371	0.303	0.762	-0.615	0.839	With SBP/DBP/HR
Heart FailureTRUE	-0.069	0.388	-0.178	0.858	-0.830	0.691	Without SBP/DBP/HR
alcoholintakegpd	0.001	0.003	0.328	0.743	-0.004	0.006	With SBP/DBP/HR
alcoholintakegpd	0.006	0.003	2.344	0.019	0.001	0.011	Without SBP/DBP/HR
bmi	-0.147	0.014	-10.520	0.000	-0.174	-0.119	With SBP/DBP/HR
bmi	-0.001	0.014	-0.064	0.949	-0.029	0.027	Without SBP/DBP/HR
CADTRUE	0.463	0.178	2.600	0.009	0.114	0.812	With SBP/DBP/HR
CADTRUE	-0.423	0.185	-2.284	0.022	-0.785	-0.060	Without SBP/DBP/HR
dbp	0.087	0.006	13.980	0.000	0.075	0.099	With SBP/DBP/HR
HypercholTRUE	-0.139	0.112	-1.242	0.214	-0.358	0.080	With SBP/DBP/HR
HypercholTRUE	-0.215	0.117	-1.838	0.066	-0.444	0.014	Without SBP/DBP/HR
HypertensionTRUE	0.663	0.107	6.226	0.000	0.454	0.872	With SBP/DBP/HR
HypertensionTRUE	1.894	0.108	17.524	0.000	1.682	2.106	Without SBP/DBP/HR
ObesityTRUE	-0.257	0.148	-1.732	0.083	-0.547	0.034	With SBP/DBP/HR
ObesityTRUE	-0.733	0.155	-4.739	0.000	-1.037	-0.430	Without SBP/DBP/HR
packyrs	0.023	0.004	6.049	0.000	0.015	0.030	With SBP/DBP/HR
packyrs	0.022	0.004	5.473	0.000	0.014	0.029	Without SBP/DBP/HR
poly(Age, 2)1	-68.774	8.357	-8.230	0.000	-85.154	-52.394	With SBP/DBP/HR
poly(Age, 2)1	-24.893	7.937	-3.136	0.002	-40.449	-9.336	Without SBP/DBP/HR
poly(Age, 2)2	-106.492	7.248	-14.693	0.000	-120.698	-92.286	With SBP/DBP/HR
poly(Age, 2)2	-113.511	7.563	-15.008	0.000	-128.336	-98.687	Without SBP/DBP/HR
PriorMed	-0.044	0.168	-0.260	0.795	-0.373	0.286	With SBP/DBP/HR
PriorMed	-0.420	0.176	-2.390	0.017	-0.764	-0.075	Without SBP/DBP/HR
pulse_rate	0.110	0.004	27.843	0.000	0.102	0.118	With SBP/DBP/HR
sbp	0.067	0.003	19.621	0.000	0.061	0.074	With SBP/DBP/HR
SexMale	0.452	0.095	4.766	0.000	0.266	0.637	With SBP/DBP/HR
SexMale	0.533	0.096	5.551	0.000	0.345	0.722	Without SBP/DBP/HR

get_pretty_fit_forest_plot(ACEi.fits.df) ->p
p

ACEi.fits.df %>% 
  get_fit_glance

generate_table.self.reported(ACEi.df, "ACEi")

Characteristic	ACEi, N = 2,4121	No ACEi, N = 25,1341	p-value2
Cardiac Age gap	0.7 (-4.6, 5.9)	0.1 (-5.4, 5.3)	<0.001
Sex			<0.001
Female	764 (32%)	13,073 (52%)
Male	1,648 (68%)	12,061 (48%)
Age	67 (61, 71)	64 (58, 70)	<0.001
Packs year of smoking	0 (0, 7)	0 (0, 3)	<0.001
Alcohol is g/day consumed	4 (0, 18)	4 (0, 15)	0.004
Hypertension	2,075 (86%)	7,123 (28%)	<0.001
Obesity	776 (32%)	4,899 (19%)	<0.001
Diabetes Mellitus	446 (18%)	1,537 (6.1%)	<0.001
Coronary Artery Disease	621 (26%)	1,487 (5.9%)	<0.001
Hypercholesterolemia	1,086 (45%)	5,205 (21%)	<0.001
Heart Failure	145 (6.0%)	267 (1.1%)	<0.001
dbp	80 (72, 86)	78 (72, 85)	<0.001
sbp	142 (131, 154)	137 (126, 150)	<0.001
pulse_rate	68 (60, 76)	68 (61, 76)	0.6
bmi	27.9 (25.1, 30.9)	26.1 (23.7, 29.1)	<0.001
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Pearson's Chi-squared test

ACEi.df %>% 
  get_stratified_age_gap_histogram("ACEi")

ACEi.df %>% 
   get_stratified_age_gap_quantiles("ACEi")

0.8.1 Lasso

ACEi.glmnet.cv <- fits.glm[["ACEi"]]

plot(ACEi.glmnet.cv) 

coef(ACEi.glmnet.cv, c(ACEi.glmnet.cv$lambda.min,
                   ACEi.glmnet.cv$lambda.1se))

## 17 x 2 sparse Matrix of class "dgCMatrix"
##                                    s1            s2
## (Intercept)             -2.286956e-16 -9.385681e-17
## SID                      8.920302e-04  .           
## SexMale                  2.276959e-02  .           
## Age                     -4.767750e-02  .           
## packyrs                  3.479136e-02  8.410991e-04
## alcoholintakegpd         2.885660e-03  .           
## PriorMed                 .             .           
## HypertensionTRUE         3.907782e-02  7.401584e-03
## ObesityTRUE             -9.937072e-03  .           
## `Diabetes Mellitus`TRUE  2.011325e-02  .           
## CADTRUE                  1.147156e-02  .           
## HypercholTRUE           -5.361039e-03  .           
## `Heart Failure`TRUE      .             .           
## bmi                     -8.367754e-02 -2.634601e-02
## sbp                      1.545082e-01  1.151769e-01
## dbp                      1.233096e-01  1.201523e-01
## pulse_rate               1.643223e-01  1.247856e-01

0.9 ARB

ARB.df <- dfs %>% 
  filter(Drug=="ARB") %>% 
  select(-Drug)

ARB.fits.df <- fits.df %>% 
  filter(Drug=="ARB")

get_pretty_fit_table(ARB.fits.df, "ARB")

ARB: Coefficients from linear models

term	estimate	std.error	statistic	p.value	conf.low	conf.high	Adjustment
(Intercept)	-20.512	0.517	-39.681	0.000	-21.525	-19.498	With SBP/DBP/HR
(Intercept)	-0.862	0.359	-2.397	0.017	-1.566	-0.157	Without SBP/DBP/HR
Diabetes MellitusTRUE	0.656	0.177	3.718	0.000	0.310	1.002	With SBP/DBP/HR
Diabetes MellitusTRUE	0.607	0.184	3.309	0.001	0.248	0.967	Without SBP/DBP/HR
Heart FailureTRUE	0.099	0.370	0.267	0.789	-0.627	0.825	With SBP/DBP/HR
Heart FailureTRUE	-0.123	0.388	-0.317	0.751	-0.882	0.637	Without SBP/DBP/HR
alcoholintakegpd	0.001	0.003	0.325	0.745	-0.004	0.006	With SBP/DBP/HR
alcoholintakegpd	0.006	0.003	2.333	0.020	0.001	0.011	Without SBP/DBP/HR
bmi	-0.147	0.014	-10.559	0.000	-0.175	-0.120	With SBP/DBP/HR
bmi	-0.001	0.014	-0.062	0.951	-0.029	0.027	Without SBP/DBP/HR
CADTRUE	0.450	0.176	2.550	0.011	0.104	0.796	With SBP/DBP/HR
CADTRUE	-0.483	0.183	-2.632	0.008	-0.842	-0.123	Without SBP/DBP/HR
dbp	0.087	0.006	14.004	0.000	0.075	0.100	With SBP/DBP/HR
HypercholTRUE	-0.143	0.112	-1.275	0.202	-0.362	0.077	With SBP/DBP/HR
HypercholTRUE	-0.226	0.117	-1.934	0.053	-0.455	0.003	Without SBP/DBP/HR
HypertensionTRUE	0.628	0.106	5.942	0.000	0.421	0.836	With SBP/DBP/HR
HypertensionTRUE	1.827	0.107	17.005	0.000	1.616	2.037	Without SBP/DBP/HR
ObesityTRUE	-0.255	0.148	-1.723	0.085	-0.546	0.035	With SBP/DBP/HR
ObesityTRUE	-0.738	0.155	-4.770	0.000	-1.041	-0.435	Without SBP/DBP/HR
packyrs	0.023	0.004	6.059	0.000	0.016	0.030	With SBP/DBP/HR
packyrs	0.022	0.004	5.456	0.000	0.014	0.029	Without SBP/DBP/HR
poly(Age, 2)1	-68.999	8.360	-8.253	0.000	-85.386	-52.612	With SBP/DBP/HR
poly(Age, 2)1	-24.507	7.942	-3.086	0.002	-40.073	-8.940	Without SBP/DBP/HR
poly(Age, 2)2	-106.467	7.247	-14.692	0.000	-120.671	-92.263	With SBP/DBP/HR
poly(Age, 2)2	-113.256	7.563	-14.974	0.000	-128.080	-98.431	Without SBP/DBP/HR
PriorMed	0.183	0.195	0.937	0.349	-0.200	0.565	With SBP/DBP/HR
PriorMed	-0.059	0.204	-0.289	0.772	-0.459	0.341	Without SBP/DBP/HR
pulse_rate	0.110	0.004	27.848	0.000	0.103	0.118	With SBP/DBP/HR
sbp	0.067	0.003	19.632	0.000	0.061	0.074	With SBP/DBP/HR
SexMale	0.452	0.095	4.772	0.000	0.266	0.637	With SBP/DBP/HR
SexMale	0.521	0.096	5.432	0.000	0.333	0.710	Without SBP/DBP/HR

get_pretty_fit_forest_plot(ARB.fits.df) ->p
p

ARB.fits.df %>% 
  get_fit_glance

generate_table.self.reported(ARB.df, "ARB")

Characteristic	ARB, N = 1,6421	No ARB, N = 25,9041	p-value2
Cardiac Age gap	1.2 (-4.1, 6.3)	0.1 (-5.4, 5.3)	<0.001
Sex			<0.001
Female	712 (43%)	13,125 (51%)
Male	930 (57%)	12,779 (49%)
Age	68 (62, 72)	64 (58, 70)	<0.001
Packs year of smoking	0 (0, 4)	0 (0, 3)	0.4
Alcohol is g/day consumed	4 (0, 18)	4 (0, 15)	0.063
Hypertension	1,535 (93%)	7,663 (30%)	<0.001
Obesity	568 (35%)	5,107 (20%)	<0.001
Diabetes Mellitus	315 (19%)	1,668 (6.4%)	<0.001
Coronary Artery Disease	295 (18%)	1,813 (7.0%)	<0.001
Hypercholesterolemia	708 (43%)	5,583 (22%)	<0.001
Heart Failure	67 (4.1%)	345 (1.3%)	<0.001
dbp	80 (74, 87)	78 (72, 85)	<0.001
sbp	144 (134, 156)	137 (126, 150)	<0.001
pulse_rate	69 (62, 78)	68 (60, 76)	<0.001
bmi	28.3 (25.9, 31.4)	26.1 (23.7, 29.1)	<0.001
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Pearson's Chi-squared test

ARB.df %>% 
  get_stratified_age_gap_histogram("ARB")

ARB.df %>% 
    get_stratified_age_gap_quantiles("ARB")

0.9.1 Lasso

ARB.glmnet.cv <- fits.glm[["ARB"]]

plot(ARB.glmnet.cv) 

coef(ARB.glmnet.cv, c(ARB.glmnet.cv$lambda.min,
                   ARB.glmnet.cv$lambda.1se))

## 17 x 2 sparse Matrix of class "dgCMatrix"
##                                    s1            s2
## (Intercept)             -2.314783e-16 -1.080653e-16
## SID                      1.962549e-03  .           
## SexMale                  2.374022e-02  .           
## Age                     -4.971222e-02  .           
## packyrs                  3.580341e-02  7.207680e-03
## alcoholintakegpd         3.515135e-03  .           
## PriorMed                 4.828414e-03  .           
## HypertensionTRUE         3.879835e-02  1.273594e-02
## ObesityTRUE             -1.039479e-02  .           
## `Diabetes Mellitus`TRUE  2.102259e-02  .           
## CADTRUE                  1.259680e-02  .           
## HypercholTRUE           -7.020339e-03  .           
## `Heart Failure`TRUE      .             .           
## bmi                     -8.573188e-02 -3.636707e-02
## sbp                      1.559629e-01  1.182037e-01
## dbp                      1.233456e-01  1.244400e-01
## pulse_rate               1.657049e-01  1.306085e-01

0.10 CCB

CCB.df <- dfs %>% 
  filter(Drug=="CCB") %>% 
  select(-Drug)

CCB.fits.df <- fits.df %>% 
  filter(Drug=="CCB")

get_pretty_fit_table(CCB.fits.df, "CCB")

CCB: Coefficients from linear models

term	estimate	std.error	statistic	p.value	conf.low	conf.high	Adjustment
(Intercept)	-20.491	0.517	-39.635	0.000	-21.505	-19.478	With SBP/DBP/HR
(Intercept)	-0.871	0.359	-2.427	0.015	-1.575	-0.168	Without SBP/DBP/HR
Diabetes MellitusTRUE	0.668	0.176	3.786	0.000	0.322	1.013	With SBP/DBP/HR
Diabetes MellitusTRUE	0.616	0.183	3.359	0.001	0.256	0.975	Without SBP/DBP/HR
Heart FailureTRUE	0.090	0.370	0.242	0.808	-0.636	0.816	With SBP/DBP/HR
Heart FailureTRUE	-0.158	0.387	-0.407	0.684	-0.917	0.602	Without SBP/DBP/HR
alcoholintakegpd	0.001	0.003	0.356	0.722	-0.004	0.006	With SBP/DBP/HR
alcoholintakegpd	0.006	0.003	2.383	0.017	0.001	0.011	Without SBP/DBP/HR
bmi	-0.146	0.014	-10.485	0.000	-0.173	-0.119	With SBP/DBP/HR
bmi	-0.001	0.014	-0.043	0.965	-0.029	0.027	Without SBP/DBP/HR
CADTRUE	0.450	0.176	2.553	0.011	0.105	0.796	With SBP/DBP/HR
CADTRUE	-0.496	0.183	-2.704	0.007	-0.855	-0.136	Without SBP/DBP/HR
dbp	0.087	0.006	13.866	0.000	0.074	0.099	With SBP/DBP/HR
HypercholTRUE	-0.137	0.112	-1.225	0.221	-0.356	0.082	With SBP/DBP/HR
HypercholTRUE	-0.219	0.117	-1.873	0.061	-0.448	0.010	Without SBP/DBP/HR
HypertensionTRUE	0.749	0.111	6.726	0.000	0.531	0.967	With SBP/DBP/HR
HypertensionTRUE	2.003	0.113	17.738	0.000	1.781	2.224	Without SBP/DBP/HR
ObesityTRUE	-0.259	0.148	-1.750	0.080	-0.550	0.031	With SBP/DBP/HR
ObesityTRUE	-0.740	0.155	-4.782	0.000	-1.043	-0.437	Without SBP/DBP/HR
packyrs	0.023	0.004	6.040	0.000	0.015	0.030	With SBP/DBP/HR
packyrs	0.022	0.004	5.461	0.000	0.014	0.029	Without SBP/DBP/HR
poly(Age, 2)1	-67.735	8.370	-8.092	0.000	-84.141	-51.329	With SBP/DBP/HR
poly(Age, 2)1	-22.451	7.952	-2.823	0.005	-38.037	-6.865	Without SBP/DBP/HR
poly(Age, 2)2	-106.555	7.247	-14.704	0.000	-120.759	-92.352	With SBP/DBP/HR
poly(Age, 2)2	-113.337	7.561	-14.989	0.000	-128.157	-98.517	Without SBP/DBP/HR
PriorMed	-0.337	0.162	-2.079	0.038	-0.655	-0.019	With SBP/DBP/HR
PriorMed	-0.685	0.169	-4.051	0.000	-1.017	-0.354	Without SBP/DBP/HR
pulse_rate	0.111	0.004	27.901	0.000	0.103	0.118	With SBP/DBP/HR
sbp	0.067	0.003	19.609	0.000	0.061	0.074	With SBP/DBP/HR
SexMale	0.461	0.095	4.863	0.000	0.275	0.647	With SBP/DBP/HR
SexMale	0.536	0.096	5.587	0.000	0.348	0.725	Without SBP/DBP/HR

get_pretty_fit_forest_plot(CCB.fits.df) ->p
p

CCB.fits.df %>% 
  get_fit_glance

generate_table.self.reported(CCB.df, "CCB")

Characteristic	CCB, N = 2,7741	No CCB, N = 24,7721	p-value2
Cardiac Age gap	0.9 (-4.0, 5.8)	0.1 (-5.4, 5.3)	<0.001
Sex			<0.001
Female	1,017 (37%)	12,820 (52%)
Male	1,757 (63%)	11,952 (48%)
Age	68 (63, 72)	64 (58, 69)	<0.001
Packs year of smoking	0 (0, 6)	0 (0, 3)	0.017
Alcohol is g/day consumed	5 (0, 19)	3 (0, 15)	<0.001
Hypertension	2,657 (96%)	6,541 (26%)	<0.001
Obesity	822 (30%)	4,853 (20%)	<0.001
Diabetes Mellitus	390 (14%)	1,593 (6.4%)	<0.001
Coronary Artery Disease	328 (12%)	1,780 (7.2%)	<0.001
Hypercholesterolemia	1,109 (40%)	5,182 (21%)	<0.001
Heart Failure	48 (1.7%)	364 (1.5%)	0.3
dbp	80 (74, 86)	78 (72, 85)	<0.001
sbp	144 (134, 155)	137 (125, 150)	<0.001
pulse_rate	70 (62, 78)	68 (60, 76)	<0.001
bmi	27.7 (25.0, 30.7)	26.1 (23.7, 29.1)	<0.001
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Pearson's Chi-squared test

CCB.df %>% 
  get_stratified_age_gap_histogram("CCB")

CCB.df %>% 
  get_stratified_age_gap_quantiles("CCB")

0.10.1 Lasso

CCB.glmnet.cv <- fits.glm[["CCB"]]

plot(CCB.glmnet.cv) 

coef(CCB.glmnet.cv, c(CCB.glmnet.cv$lambda.min,
                   CCB.glmnet.cv$lambda.1se))

## 17 x 2 sparse Matrix of class "dgCMatrix"
##                                    s1            s2
## (Intercept)             -2.295597e-16 -1.213257e-16
## SID                      1.521459e-03  .           
## SexMale                  2.394914e-02  1.778284e-03
## Age                     -4.834152e-02 -4.750819e-03
## packyrs                  3.542791e-02  1.261464e-02
## alcoholintakegpd         3.484559e-03  .           
## PriorMed                -1.058849e-02  .           
## HypertensionTRUE         4.446964e-02  1.773942e-02
## ObesityTRUE             -1.040865e-02  .           
## `Diabetes Mellitus`TRUE  2.099499e-02  .           
## CADTRUE                  1.218835e-02  .           
## HypercholTRUE           -6.316453e-03  .           
## `Heart Failure`TRUE      .             .           
## bmi                     -8.464660e-02 -4.518251e-02
## sbp                      1.553599e-01  1.228784e-01
## dbp                      1.225816e-01  1.258928e-01
## pulse_rate               1.656166e-01  1.360434e-01

0.11 Diuretics

diuretics.df <- dfs %>% 
  filter(Drug=="Diuretics") %>% 
  select(-Drug)

diuretics.fits.df <- fits.df %>% 
  filter(Drug=="Diuretics")

get_pretty_fit_table(diuretics.fits.df, "Diuretics")

Diuretics: Coefficients from linear models

term	estimate	std.error	statistic	p.value	conf.low	conf.high	Adjustment
(Intercept)	-20.518	0.517	-39.695	0.000	-21.531	-19.505	With SBP/DBP/HR
(Intercept)	-0.874	0.359	-2.433	0.015	-1.578	-0.170	Without SBP/DBP/HR
Diabetes MellitusTRUE	0.664	0.176	3.763	0.000	0.318	1.009	With SBP/DBP/HR
Diabetes MellitusTRUE	0.608	0.183	3.319	0.001	0.249	0.968	Without SBP/DBP/HR
Heart FailureTRUE	0.093	0.371	0.249	0.803	-0.635	0.820	With SBP/DBP/HR
Heart FailureTRUE	-0.089	0.388	-0.230	0.818	-0.851	0.672	Without SBP/DBP/HR
alcoholintakegpd	0.001	0.003	0.329	0.742	-0.004	0.006	With SBP/DBP/HR
alcoholintakegpd	0.006	0.003	2.323	0.020	0.001	0.011	Without SBP/DBP/HR
bmi	-0.147	0.014	-10.537	0.000	-0.174	-0.120	With SBP/DBP/HR
bmi	0.000	0.014	-0.022	0.982	-0.028	0.028	Without SBP/DBP/HR
CADTRUE	0.456	0.176	2.585	0.010	0.110	0.801	With SBP/DBP/HR
CADTRUE	-0.483	0.183	-2.634	0.008	-0.842	-0.123	Without SBP/DBP/HR
dbp	0.087	0.006	13.991	0.000	0.075	0.100	With SBP/DBP/HR
HypercholTRUE	-0.142	0.112	-1.269	0.204	-0.361	0.077	With SBP/DBP/HR
HypercholTRUE	-0.222	0.117	-1.900	0.057	-0.452	0.007	Without SBP/DBP/HR
HypertensionTRUE	0.643	0.105	6.133	0.000	0.437	0.848	With SBP/DBP/HR
HypertensionTRUE	1.849	0.106	17.403	0.000	1.641	2.058	Without SBP/DBP/HR
ObesityTRUE	-0.257	0.148	-1.733	0.083	-0.547	0.034	With SBP/DBP/HR
ObesityTRUE	-0.738	0.155	-4.768	0.000	-1.041	-0.435	Without SBP/DBP/HR
packyrs	0.023	0.004	6.045	0.000	0.015	0.030	With SBP/DBP/HR
packyrs	0.022	0.004	5.468	0.000	0.014	0.029	Without SBP/DBP/HR
poly(Age, 2)1	-68.955	8.366	-8.242	0.000	-85.352	-52.557	With SBP/DBP/HR
poly(Age, 2)1	-24.164	7.944	-3.042	0.002	-39.734	-8.595	Without SBP/DBP/HR
poly(Age, 2)2	-106.461	7.247	-14.691	0.000	-120.666	-92.257	With SBP/DBP/HR
poly(Age, 2)2	-113.260	7.563	-14.975	0.000	-128.084	-98.436	Without SBP/DBP/HR
PriorMed	0.113	0.221	0.511	0.610	-0.320	0.545	With SBP/DBP/HR
PriorMed	-0.289	0.231	-1.254	0.210	-0.741	0.163	Without SBP/DBP/HR
pulse_rate	0.110	0.004	27.846	0.000	0.103	0.118	With SBP/DBP/HR
sbp	0.067	0.003	19.637	0.000	0.061	0.074	With SBP/DBP/HR
SexMale	0.452	0.095	4.774	0.000	0.266	0.638	With SBP/DBP/HR
SexMale	0.517	0.096	5.381	0.000	0.329	0.705	Without SBP/DBP/HR

get_pretty_fit_forest_plot(diuretics.fits.df) ->p
p

diuretics.fits.df %>% 
  get_fit_glance

generate_table.self.reported(diuretics.df, "diuretics")

Characteristic	diuretics, N = 1,2301	No diuretics, N = 26,3161	p-value2
Cardiac Age gap	1.0 (-3.9, 5.8)	0.1 (-5.4, 5.3)	<0.001
Sex			0.6
Female	608 (49%)	13,229 (50%)
Male	622 (51%)	13,087 (50%)
Age	68 (63, 72)	64 (58, 70)	<0.001
Packs year of smoking	0 (0, 6)	0 (0, 3)	0.019
Alcohol is g/day consumed	3 (0, 14)	4 (0, 15)	0.003
Hypertension	1,148 (93%)	8,050 (31%)	<0.001
Obesity	447 (36%)	5,228 (20%)	<0.001
Diabetes Mellitus	202 (16%)	1,781 (6.8%)	<0.001
Coronary Artery Disease	202 (16%)	1,906 (7.2%)	<0.001
Hypercholesterolemia	546 (44%)	5,745 (22%)	<0.001
Heart Failure	83 (6.7%)	329 (1.3%)	<0.001
dbp	80 (73, 86)	78 (72, 85)	<0.001
sbp	143 (131, 155)	138 (126, 150)	<0.001
pulse_rate	70 (62, 79)	68 (61, 76)	<0.001
bmi	28.5 (25.7, 31.8)	26.2 (23.7, 29.2)	<0.001
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Pearson's Chi-squared test

diuretics.df %>% 
  get_stratified_age_gap_histogram("Diuretics")

diuretics.df %>% 
 get_stratified_age_gap_quantiles("Diuretics")

0.11.1 Lasso

diuretics.glmnet.cv <- fits.glm[["Diuretics"]]

plot(diuretics.glmnet.cv) 

coef(diuretics.glmnet.cv, c(diuretics.glmnet.cv$lambda.min,
                   diuretics.glmnet.cv$lambda.1se))

## 17 x 2 sparse Matrix of class "dgCMatrix"
##                                    s1            s2
## (Intercept)             -2.278847e-16 -1.144220e-16
## SID                      8.914631e-04  .           
## SexMale                  2.281645e-02  .           
## Age                     -4.776635e-02 -3.901188e-04
## packyrs                  3.477911e-02  1.000863e-02
## alcoholintakegpd         2.893804e-03  .           
## PriorMed                 1.147081e-03  .           
## HypertensionTRUE         3.879576e-02  1.517539e-02
## ObesityTRUE             -9.901109e-03  .           
## `Diabetes Mellitus`TRUE  2.008850e-02  .           
## CADTRUE                  1.145429e-02  .           
## HypercholTRUE           -5.394407e-03  .           
## `Heart Failure`TRUE      .             .           
## bmi                     -8.377949e-02 -4.076007e-02
## sbp                      1.545863e-01  1.195392e-01
## dbp                      1.232900e-01  1.262449e-01
## pulse_rate               1.643293e-01  1.331395e-01

0.12 Self-reported drugs used

read_csv("../data/Derived/self_reported_drugs_used.csv")

## Rows: 79 Columns: 2
## ── Column specification ──────────────────────────────────────────────────────────────────────────────
## Delimiter: ","
## chr (2): meaning, Drug
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.